How an AI‑Powered IDE Accelerated the 3‑Day Build of an English Learning App
In just three days, the team built the Streams‑to‑River English learning app by leveraging the AI‑driven TRAE IDE, which generated most of the code, integrated large‑model features, and employed a Hertz‑Kitex microservice architecture with MySQL, Redis, and AI services.
The author, ByteDance R&D leader Hong Dingkun, presented the English‑learning app “Streams to River” (experience address: https://sstr.trae.com.cn) at the Volcano Engine Force conference.
The app records, extracts, and manages everyday English words, sentences, and contexts, applying the Ebbinghaus forgetting curve for periodic review and retention.
Approximately 85% of the code was generated by AI through natural‑language dialogue, thanks to the TRAE AI IDE, which enabled rapid development and debugging despite limited dedicated time.
TRAE was extensively used for code creation, debugging, commenting, and unit‑test writing, and its workflow integrated image‑to‑text, real‑time chat, speech recognition, and word‑highlighting capabilities from large models.
The team plans to open‑source the entire codebase on GitHub for developers to explore the collaboration between TRAE and human developers. All code is available at https://github.com/Trae-AI/stream-to-river.
Project Information
Streams to River is built on the Hertz and Kitex frameworks as a microservice system for word learning and language processing, offering a full solution from API services to RPC implementation, including user authentication, word management, review tracking, real‑time chat, speech recognition, and image‑to‑text modules.
The system follows a front‑back separation architecture with four layers:
API service layer: HTTP APIs based on Hertz, handling front‑end requests.
RPC service layer: Business logic implemented with Kitex, serving API layer calls.
Data access layer: MySQL database and Redis cache for persistence and caching.
Intelligent processing layer: Integration of large language models (LLM), automatic speech recognition (ASR), and image‑to‑text functions.
AI Coding Practice Insights
TRAE’s core features “code completion” and “local code generation” automatically infer and fill code based on context, boosting programming efficiency.
The “natural language programming” approach does not replace engineers; a 300‑line feature can be described in about 200 characters, yet the implementation remains an engineering effort.
The underlying model, doubao‑dev, built on ByteDance’s latest Doubao 1.6 series, was further fine‑tuned for engineering scenarios, enhancing development productivity.
About TRAE
TRAE aims to be the “real AI engineer,” providing an AI IDE that seamlessly integrates into developers’ workflows to deliver higher quality and efficiency.
Try TRAE at https://www.trae.cn/.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
Rare Earth Juejin Tech Community
Juejin, a tech community that helps developers grow.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.